-----Original Message-----
From: Eric Blake [mailto:eblake@redhat.com]
Sent: Tuesday, October 22, 2013 2:23 PM
To: Chen Hanxiao; 'Michal Privoznik'
Cc: libvir-list(a)redhat.com
Subject: Re: [libvirt] [PATCH 5/5]virsh: mark '--shareable' as
VSH_OFLAG_IGNORE
On 10/22/2013 07:13 AM, Chen Hanxiao wrote:
>>>
>>
>> Wow, we've never wanted to hide any option before? I've recall
>> introducing VSH_OT_ALIAS, but there hasn't been anything else?
We've always hidden one spelling that duplicates another, but this is a
case where a boolean option is a duplicate of a non-boolean option, so
I'm not sure if VSH_OT_ALIAS alone does what we want (or maybe it just
means we enhance VSH_OT_ALIAS to be able to express a boolean alias for
a non-boolean default).
>
> Do you mean we just treat --shareable as an alias for --mode like this:
>
> {.name = "shareable",
> - .type = VSH_OT_BOOL,
> + .type = VSH_OT_ALIAS,
> - .help = N_("shareable between domains")
> + .help = "mode"
If it works. Maybe even by having it be an alias for .help =
"mode=shareable" rather than just "mode", and teach VSH_OT_ALIAS to
handle '=' within the .help text.
Better choice.
> },
>
> But with VSH_OT_ALIAS, auto complete could also get it.
Command line completion is orthogonal. It needs fixing anyways (there
were patches posted last month, but I haven't seen any recent action on
them).
Ideally, we want the following:
attach-disk <TAB>
to NOT show --shareable (because it can show --mode instead)
attach-disk --sh <TAB>
would show --shareable (just because it's deprecated and hidden by
default does NOT mean it must be hidden if the user explicitly requests
something that can't complete in any other way)
attach-disk --mode=shareable --sh<TAB>
attach-disk --mode readonly --sh<TAB>
would both complain about no possible completion (because --shareable
has already been eliminated by --mode appearing earlier).
attach-disk --shareable --mo<TAB>
would complain about no possible completion (--shareable implies
--mode=shareable, and --mode can't appear more than once).
But getting that working in this patch series is NOT a prerequisite -
save it for everything else that needs fixing with completion.
Thanks for your elaborate explanation.
I'll try to reach these goals, although they're a little complex.
> We still need new flag to hide it.
I'm still not convinced we need VSH_OFLAG_IGNORE. VSH_OT_ALIAS, plus
new logic that lets a boolean be an alias to non-bool=value, ought to be
the way to do it (ie. fix our existing aliasing mechanism, rather than
adding a second aliasing mechanism).
Agree.
I'll try to expand the logic of VSH_OT_ALIAS.
Thanks!
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library
http://libvirt.org